This page is for VC platform leads who run several communities, an LP circle, a founder alumni group, an operator guild, and want to know which ones turn into CEO conversations that lead somewhere, without turning their members into a scoreboard.
Why the member is the wrong unit
A platform team decides at the network level: which community to brief, where to spend program hours, which group needs better material. Those decisions need network comparisons. Member scores answer a question nobody should be asking, and asking it does damage.
- Tiny counts. A member's handful of asks a year cannot separate skill from luck.
- Rewarding the wrong habit. The program rests on the CEO's yes. SourceX's screening page treats an introduction the decision maker never heard about as usually a poor fit. A member who decides not to ask a CEO who would not welcome it would rank below one who asks everyone, so a ranking punishes the judgment the program depends on.
- Exposed relationships. A member table shows who knows which CEO and whose asks failed, information members logged for the platform's learning, not for display.
- Dishonest logs. Members who expect to be ranked log selectively, and the comparison collapses with them.
SourceX's own terms keep partners apart. Section 7 says, "Partners never receive access to other partners' records, applicant documents or confidential buyer information." A platform that pools its members' logs should hold itself to at least that separation.
| Term | How this page uses it |
|---|---|
| Network | A group the platform already runs, with a membership list fixed for the quarter |
| Ask | One member asking one CEO whether they would like a conversation about SourceX |
| Active member | A consenting member who made at least one ask in the quarter |
| Steward | The one person who holds the key linking members to codes; not an analyst and not someone who sets network budgets |
| Qualified outcome | A CEO's written yes followed by a verified application inside the 90-day window, with no published fit fact failing |
The privacy-preserving cohort design
Each layer closes one route to a member ranking. The numbers are study settings, labeled Illustrative, and they are fixed before the quarter begins.
| Layer | Rule | What it prevents |
|---|---|---|
| 1. Consent | Members agree to network-level counting, are told no member ranking exists or will exist, and may withdraw their asks before the read date | Logging under false expectations |
| 2. Logging | Each member logs the date, the network, the stages read from their own dashboard and a company code of their choosing; never a company name, record or figure | Exposing companies |
| 3. Key file | The steward alone holds the member-to-code table, kept apart from the logs and deleted after the read date | Analysts learning who asked whom |
| 4. Contribution cap | The steward keeps the first three asks per member per quarter, in date order; later asks stay in the member's own log | One prolific member defining a network, or a network figure being reworked into one person's |
| 5. Analysis file | Network, quarter and outcome stage; no member field at all | A ranking being computed later |
| 6. Size floors | A network is reported only with eight or more active members and twenty or more asks in the sample | A "network" figure that is really one or two people |
| 7. Suppression | Any cell from 1 to 4 shows as "fewer than 5", and a second cell is withheld wherever a total would let a reader recover it | Recovering a small count by subtraction |
| 8. Member view | A member may ask the steward for their own counts; never anyone else's, never a position | Leaderboards assembled on request |
Layer 7 borrows its logic from health data. ResDAC's summary of the CMS cell-size policy says: "No cell can be reported that allows a value of 1 to 10 to be derived from other reported cells or information." That threshold suits large health files. Counts that stay inside the platform team here use the lower threshold in the table; anything shown outside the team waits until its counts can meet a stricter published rule like that one.
Running the comparison
Every network gets the same ladder, read on day 120 after the quarter's last ask so each row has passed its 90-day window: asks; CEO agreed; verified application; qualified outcome; CEO declined; no reply; duplicate under section 3; still open. A network with fewer than twenty asks in the sample shows counts, never shares.
Members may also log "considered, not asked" when they decide a company or a moment is wrong. That count is reported on its own line and never counted against the network. A network whose members leave unsuitable CEOs alone is doing its job.
The comparison is descriptive. Networks differ in who joins them and which companies their members know, so a gap between two networks is an association, not proof that one community works better. For a like-for-like version, the matched-pair design in do specialized networks produce better-qualified referrals pairs asks about similar companies; all eight layers here apply to it unchanged.
How the rules act on one quarter (Illustrative)
Invented counts, shown only to make the rules visible. They are not findings.
| Network | Active members | Asks in sample | CEO agreed | Agreed, then verified application | Agreed, no application yet | Declined or no reply | Reported? |
|---|---|---|---|---|---|---|---|
| LP circle | 9 | 22 | 6 | Fewer than 5 | Withheld | 16 | Yes, with two cells masked |
| Founder alumni | 14 | 31 | 12 | 6 | 6 | 19 | Yes, nothing masked |
| Operator guild | 5 | 11 | Withheld | Withheld | Withheld | Withheld | No: below eight active members |
The LP circle's verified applications fall below five, so that cell is masked; its "agreed, no application yet" cell is withheld as well, because six agreed minus that cell would reveal the masked one. The founder alumni row clears every rule and prints in full. The guild misses the floor and disappears entirely, and so does any all-networks total row, since subtracting the two reported networks from it would expose the guild.
Missing, declined and unsuccessful asks
- Every ask in the sample stays: declined, unanswered, duplicate, declined in SourceX's review or still open on the read date. Open is its own row, never a loss.
- Enrolled members who never asked count toward the network's membership, so a network with many silent members shows it.
- Withdrawals are removed before analysis, and only their total across all networks is published.
- Asks above the cap are reported as one total across all networks, never by network, so the figure cannot point to a busy member.
- Small samples: with twenty or thirty asks per network, only large gaps can show. "No difference detectable" is a result and is published as one, alongside any network that never reached the floor.
Risks and limits of the design
- The steward is a single point of trust. If the steward also reviews members' performance, the design fails. Choose someone with no say over members.
- The cap discards information. A network whose value comes from a few connectors will look weaker than it is; the network concentration report studies that dependency directly, under its own consent.
- Consent skews the logs toward keen members, and a network where many decline to log may look better or worse than it is.
- Review and program changes sit outside the network. SourceX decides rights and program fit in its own review, and its screening page says criteria follow each program and can move; a quarter-to-quarter swing may reflect a program change, not the network.
- Pressure to rank anyway. The analysis file cannot produce a member list, which is the point; the request should still be refused in writing.
- Small networks may never qualify for comparison. "Not reported" is the honest output for them.
Publishing anything beyond the team adds company-level risks this page does not cover; the twelve-gate review in publishing network research without exposing companies handles those, and the office-by-office reporting guide shows how to keep context beside each network's row.
Methods note. The layers rest on SourceX's terms and screening page and on the CMS cell-size rule as ResDAC describes it, all checked on 2026-10-28. The floors of eight members and twenty asks, the cap of three asks a quarter and the under-five threshold are our study settings, untested. Not measured: the value of any company's records, payouts, SourceX's review decisions or any member's individual results. For a scorecard that tracks one network over time, see what would prove a venture network is becoming more useful.
What every network's asks must keep true
A comparison is only as sound as the asks inside it. A company needs 50 or more full-time employees at peak, without counting contractors; years of records it created; rights to license them; and a CEO who agrees to the conversation and has authority to start it. Funding plays no part. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English. Members ask about the kind of records a company keeps at most; files, samples and logins never leave it.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
Set the eight layers before the quarter opens. When a member's ask turns into a CEO's yes, that member can register, weigh the company against who qualifies and read the program terms before submitting.
Comparing networks is referral work; deciding whether SourceX deserves investment is a separate exercise, and materials for it are available on request from the SourceX investor overview.
This article is for information only and is not an offer to sell or a solicitation of an offer to buy securities. Any offering by SourceX is made only to verified accredited investors through its offering documents.